Most offline setups resemble mixing boards more than anything else. Or that's what they're best at atm. Online portals focus more on random image generation with tight bounds and as a result you get something decent most of the time.

From random reddit stuff this is where the current models shine...

https://www.reddit.com/r/StableDiffu...op_and_stable/

...simple sketch that's locked with a control net and then you can just shotgun different rendering variations. You need something resembling taste, but rendering is the part that consumes most time when you do stuff by hand. You can still do editing based variants at a reasonable pace, but anything more complicated gets tricky. Clients don't know wtf they want, so revisions are standard and ability to come up with fast variants just to find out what to refine is worth money.

This was released few days ago...

https://www.youtube.com/watch?v=nJ1yQn17lbE

...french animation studio called Blue Spirit handled the animation. It's about 90% 3d animation flipped to 2d. 100% of it is ran through algos. Now how much of it is AI related really just depends on the definition. I don't personally care about the machine learning bit since it's really just a way of creating rules, it's never let run wild and everything it creates falls apart without procedural generation running on hard coded rules.

Regarding the show. Backgrounds are beautiful and character design/animation is about 80% where it should be. You get used to the uncanny valley rotoscoping effect. The show took about 10% or less of the assets/time compared to Arcane (show that itself was 95%+ done in pure 3d 100% algo good times filtering, but no one seems to care). It shows, but the scale difference regarding resources isn't evident. Writing is ass. The show is whatever, but this way of doing animation stuff will become the norm.

Oh and random quote from CEO of Blue Spirit...

https://www.vfxvoice.com/circling-th...imation-today/

...” Game engines and AI are going to have a dramatic impact on the animation pipeline and workflow, Lelardoux says. “Soon you will have the possibility to keyframe directly in the game engines. It means that the animators will be in a totally immersive way to do animation in the future. Maybe AI will be the solution for the studio and producers to create content faster and cheaper. Artists and creators have to be able to trust studios, so that’s why we have to be careful with AI.”

The quote without the PR filter is roughly, "this is close to what we're doing now, get with the times or get fucked".

Somewhat related...

https://www.youtube.com/watch?v=7ttG90raCNo

...4 part series with the first 3 released.

It's a sane look on CGI in movies. Use of practical effects has become an annoying buzzword that's used in marketing. We're approaching the sentiment of "practical good, CGI bad".

Excluding the use of AI in VFX, the other similarity is the constantly moving goalpost on what's "real" and/or "cheating". CGI and practical are both just VFX. They often support each other. Before CGI environments matte painting and miniatures did the same. Miniatures are still used sometimes, but i assume matte painting is gone. Both were considered cheating in the past. Use of stuntmen was considered cheating.

People that don't do stuff glorify wasting time and struggling. They are considered virtues on their own. That's just dumb shit.

Oh one thing that pops up multiple times when using practical effects alongside CGI is how useful they are as a lighting reference. It's math and physics. People are really bad at both of those, because they get very complicated very fast. Computers can map those, but without a reference you need to build entire scenes that are often way bigger than what's on the screen. With a reference you can approximate them more accurately.

Anyways the series goes through a dozen films that were marketed as all practical effects and in reality 90% CGI. Some history of VFX and then one recent movie that was made with only practical effects.